Abstract:  Traditional dashboard detection mainly relied on manual labor, so the efficiency and accuracy of the detection were not high. Aiming at these problems, a dashboard detection system was designed based on machine vision. Firstly, the overall composition and software of the detection system was introduced. The pointer detection algorithm technological process of the system was introduced in detail. According to whether the pointer and the background were easy to separate, car instruments were divided into two categories. Then,the threshold segmentation method was used to roughly locate the pointer. According to the type of the dashboard, the skeleton extraction and gray scale method were used to refine the pointer. To finely locate the pointer, the iterative weighted least square method based on the Tukey weight function was used to fit the pointer straight line. A method that minimized the sum of the distances between center and intersections was proposed to locate the pointer rotation center. The angle method was used to identify pointer readings. Finally, instruments A and B were tested in the detection system. The results show that the error between the system detection value and the theoretical deflection value, that is, the reading error of the system pointer detection is less than 1%, the algorithm has good accuracy. And the average recognition time of the detection system is about 200ms, which proves that the recognition accuracy and speed can meet the actual detection requirements.

Key words:  auto dashboard;pointer detection;machine vision;least squares method

HAO Yong-fei, TANG Xu-sheng, CHENG Liang-li. Auto dashboard pointer detection based on machine vision［J］.Journal of Mechanical & Electrical Engineering, 2022,39(1):134-140.